International Journal of Computer Applications (0975 8887)

Volume 27 No.6, August 2011

20
Optimum Power Loss Analysis of Radial Magnetic Bearing
using Multi-Objective Genetic Algorithm
Santosh shelke
PhD scholar
National Institute of technology, Warangal.
Andhra Pradesh (India).
R.V.Chalam
Professor
National institute of technology, Warangal
Andhra Pradesh (India)
ABSTRACT
In this paper, weight optimization of radial magnetic bearing
(RMB) for varying poles has been carried out using multi-
objective genetic algorithms (MOGAs). The total weight of
RMB and copper loss has been selected as the minimization type
objective function. The maximum space available, saturation
flux density, the maximum current densities that can be supplied
in the coil and the load to be lifted have been chosen as
constraints. The coil space radius, pole tip radius, radial length of
coil and number of poles has been proposed as design variables.
Apart from the comparison of performance parameters in the
form of figures and tables, designs are also compared through
line diagrams. Post-processing has been done on the final
optimized population by studying the variation of different
parameters with respect to objective functions. A criterion for the
choice of one of the best design based on the minimum weight of
bearing showing optimum copper loss.
Keywords
Radial Magnetic Bearings, Genetic Algorithms, Optimum
Design, Multi-Objective Optimization
1. INTRODUCTION
In magnetic bearings, the transformation of energy takes place in
two steps. First, the electrical energy transformed into the
magnetic energy, and then it transforms into the mechanical
energy. An AMB provides a non-contact means of supporting a
rotating shaft through an attractive magnetic levitation force and
hence they offer many advantages over conventional bearings.
Active magnetic bearings are typical mechatronic product. They
are composed of mechanical components combined with
electronic elements such as sensors, power amplifiers and
controllers which may be in the form of a microprocessor. As
shown in figure1.besides the controller, the general control
system also includes the sensor, A/D and D/A conversion and
power amplifier. The rotor displacement along one of the axes is
detected by these position sensors and converted into signals of
standard voltage. Then compared with the setting value, the error
signal enters the controller. After A/D conversion, the controller
processes this digital signal according to a given regulating rule
(control arithmetic) and generates a signal of current setting.
After D/A conversion, this current signal enters the power
amplifier, whose function is to maintain the current value in the
electric magnet winding at the current level set by the controller.
Therefore, if the rotor leaves its center position, the control
system will change the electromagnet current in order to change
its attraction force and, respectively, draws the rotor back to its
balance position.
Fig 1: Block diagram of AMB system
In the present paper, a multi-objective optimization procedure of
magnetic bearings have been presented and illustrated. Two
objective functions have been considered, namely the minimization
of copper loss and the minimization of weight to decide
optimum number of poles for constant electromagnetic force to
lift the rotor.
Bakay L., and Dubois M., [1] studied effect of Cu and Iron
losses of optimized eight pole radial AMB on discharge time of
no load long term flywheel energy storage. NSSN configuration
is used. It concludes that for high discharge time for low loss
AMB mass is smaller than in case of low discharge time.
Optimal solution is for class of sinusoidal force signal. Also
presented static allocation strategy for suboptimal power loss.
Bakay L., Dubois M., and Ruel J.,[2] optimized AMB to
minimize Cu and Iron loss for different magnitude of external
force. Subsequently, Bloodgood et al. [3] applied the theory for the
deterministic optimal design of a thrust magnetic bearing with
bias permanent magnets. The magnetic bearing design included
the determination of geometries of the bearing and the current
supplied to carry a specified load for a specified gap.
Optimal design was carried out in two steps, including modeling
the magnetic circuit, which determines the accuracy of
achieving the requirement, and optimization of the design,
which determines the efficiency of achieving the requirement
[3]. Optimal design by using Gas of radial active magnetic
bearings integrated with their control was studied by Chang and
Chung [4]. Many techniques of multiobjective genetic
International Journal of Computer Applications (0975 8887)
Volume 27 No.6, August 2011
21
algorithms (MOGAs) have been developed in recent years [7, 9,
11] among which Nondominated Sorting Genetic Algorithm-II
(NSGA-II) [8] is one of the widely used technique [5,10, 12]. Rao
and Tiwari [6, 13] attempted to optimize the physical parameters of
design of thrust magnetic bearings and rolling element bearing by
using single objective genetic algorithms. The design optimization
was performed by taking two objective functions, namely
minimization of the power loss and maximization of the load to
weight ratio, independently. The designs were compared for the
cases with and without permanent magnets. Various approaches
have been developed to aid the design of magnetic bearings.
Parametric and systematic approaches have been developed to
optimize the power to weight ratio of magnetic bearings [14] and
to achieve minimum power. These approaches are useful in
developing design strategies. Ying GAO, Lei Shi, Pinging
Yao[15]presented new form of MOGA,steady state non
dominated sorting genetic algorithm(SNSGA) is realized by
combining steady state idea in single objective genetic
algorithm(SOGA) and fitness assignment strategy of
nondominated sorting genetic algorithm. R.T.marler, J.S.Arora
[16] suggested that decision making can be done on the basis of
a priori or posteriori approach. In this method one needs prior
knowledge of trade-off for taking decision. The different
designs are generated as a Pareto-frontier and the designer can
take a decision of choice depending on the different trade-offs
generated. Tomoharu Nakashima, Hisao Ishibuchi [17] examines
ability of genetic algorithm to find compact reference set nearest
neighbor classification. In this case two objective are to
maximize classification of reference set and to minimize size of
reference set. Deepti Chafekar, Liang shi, Jiang Xuan [18]
proposed novel method for solving multiobjective optimization
problem using reduced models. Santosh shelke, R.V.Chalam [19,
21] proposed theoretical design of eight pole radial magnetic
bearing and copper loss optimization.E.H.maslen [20] discussed
fundamentals and optimum design of radial and thrust magnetic
bearing.
2. GEOMETRY of RADIAL MAGNETIC
BEARING
Fig 2: Radial Magnetic Bearing
Where,
g
l - Air gap,
c
t Thickness of coil,
c
l radial length of coil on
pole,
c
r -coil space radius,
j
r -Radius of journal,
r
r Radius of
rotor,
p
r Pole tip radius,
s y
w - Width of stator
yoke,
r y
w - Width of rotor yoke. Gap area, .
g p p
A w l = ,
p
w -width of pole,
p
l -length of pole,
p
n -Number of poles,
For available coil space thickness of coil is calculated as,
tan
2
p
c p
p
w
t r
n

| |
| |
= + ]
]
]
]
\
\
(1)
We know that maximum force,
max
F carried by AMB
2
max 0
( )
sat p g
F B n A = (2)
3. MULTI-OBJECTIVE OPTIMIZATION
The multi-objective optimization problem consists of objective
functions to be optimized by design variables while satisfying
certain constraints.
3.1. Bounds for Variables
Table 1. Choice of design vectors.
min max
0.3 , 0.8 ;
c c
r m r m = =
min max
0.03 , 0.08 ;
p p
r m r m = =
min max
0.2 , 0.8 ;
c c
l m l m = =
min max
4, 12;
p p
n n = =
3.2 Input Parameters
Table 2. Constant parameters [18] and inputs assumed for
radial magnetic bearing design
Permeability of vacuum,
7
0
4 10 / ; H m
-
=
Resistivity,
5
2 10 ; m
-
= O
Air gap length, 0.001 ;
g
l m =
Radius of journal, 0.03 ;
j
r m =
Width of stator , 0.01 ;
sy
w m =
Saturated flux density, 1.2 ;
sat
B T =
flux density,
min
0.2 ; B T =
Coil mmf loss factor, 1.394;
i
K =
Actuator loss factor, 1.072;
a
K =
Coil packing factor, 0.85; =
Electromagnetic force, 350 ; F N =
Maximum volume of coil,
3
max
0.00039 ; V m =
Maximum copper loss,
max
500 ; P W =
Iron saturation factor, 0.5; =
Weight densities,
3
7770 / ;
s
kg m =
3
8910 / ;
c
kg m =
Current density,
2
600000 / ;
ub
J A m =
3.3 Objective Function
r
r
w
sy
c
l
rc
g
l
ry
w
r
j
w
p
r
p
International Journal of Computer Applications (0975 8887)
Volume 27 No.6, August 2011
22
When the genetic algorithm is applied to two objective functions,
namely the copper power-loss in the magnetic coil and the weight
of the magnetic bearing, it is observed that these are mutually
conflictive [6], i.e. with the decrease in the power-loss causes an
equivalent increase in the weight and visa versa.
3.3.1 Power loss in coil
From fundamental relation in magnetic, bearing flux density in
terms of coil current is given by,
0
2
g
ni
B
l

= (3)
Where,
ni is the magneto- motence and
0
is the permeability of the
vacuum.
Groom and Bloodgood [3] proposed a model by adding coil mmf
loss factor, K
i
and actuator loss factor, K
a
to ideal model.
0
2
i
a g
K ni
B
K l

= (4)
The magnetic force exerted by the actuator could be expressed as
0
g
BA
F

=
(5)
The standard form of expression of the power-loss in the coil is given
as
2
P i R =
(6)
2
c c
P J A V =
(7)
The cross sectional area and the volume of the coil are expressed as
( )
c c c p
A t r r = - and
c c c
V A l =
(8)
3.3.2 Weight of magnetic bearing
The overall weight of the magnetic bearing could be expressed
as
Total c s
W W W = + (9)
Where,
Weight of coil, and stator is
c c c
W V = And
s s s
W V = (10)
( ) ( )
s s c sy c p g
V r r w r r A = - + -
(11)
3.4 Constrains
The maximum and minimum forces exerted by the
bearing, F
max
and F
min
, are determined from the current stiffness
and the displacement stiffness at corresponding gaps, and these
form equality constraints [3]. Recall equation (4) and (5) to
express electromagnetic force are
max, min
2
max, min
max, min 0
(0.25 )
i p
g
a g
K n i
F A
K l

| |
] =
]
\
(12)
Where,
max, min
max, min
( )
g g
l l x = (13)
Maximum displacement of rotor in terms of force and
displacement stiffness,
max max
( ) /
x
x F F K = - (14)
The current density supplied in the coil should not be more
than the maximum current density that the coil material can
sustain, which is called the coil saturation current density.
The maximum and minimum magneto-muteness required could
be found and then current densities required to be supplied in the
coil is
max, min max, min p c
n i A J = (15)
Fromequation (12) and (15) we get,
max, min max, min
max, min
0
4
a g
i c g
K l
F
J
K A A
| |
= ]
]
\
(16)
Hence, constraint becomes,
max min
; 0
sat
J J J : : (17)
The magnetic flux density in the stator-iron should not be
more than the iron magnetic saturation flux density. The maximum
and minimum magnetic flux densities of the stator that are
required to support the maximum and minimum loads are
expressed as,
max, min 0 max, min g
B F A = (18)
From above it can be observed that for a fixed maximum
force, as the pole-face area of the air-gap increases, the maximum
magnetic flux density required in the stator-iron reduces.
For a linear range of operation the actuator could be designed to
have the magnetic flux density between two limits. In this work an
iron saturation factor, , is used to express the linear range as a
percentage of the magnetic saturation flux.
According to Equation (4) , the linear range limits [4] for lower
and upper bounds are expressed as
min
0 min
min
2
i
sat
a g
K ni
B
K l

: (19)
and
max
0 max
max
2
i
sat
a g
K ni
B
K l

s (20)
Equations (19) and (20) constrain the choice of
max
ni .
The choice of
max
ni influences the cross sectional area of the coil.
As the thickness of coil increases, the available thickness of the
iron core decreases which in turn increases the magnetic flux
density for supporting a fixed load. Thus, equations (19) and (20)
influence the choice of design parameters, i.e. the inner and outer
radii of the coil.
International Journal of Computer Applications (0975 8887)
Volume 27 No.6, August 2011
23
Table 3. The summary of the formulation of the magnetic
bearing design for the multi-objective optimization
Minimize , , 1, 2
i
f x =
Where,
x =
, ,
{ , }
c p c p
r r l n
Subject to;
, , 0;
i
g x : 1, 2......, 6; j =
, , 0,
k
h x = 1, 2; k = ; 1, ...., 4.
L U
p
x x x p s s =
Where,
, , , ,
1 2
; .
cu Total
f x p f x W = =
, ,
1 max
;
sat
g x J J = -
, ,
2 min
; g x J =
, ,
3 max max
;
sat
g x B B = -
, ,
4 min min
;
sat
g x B B = -
, ,
5 max
;
c
g x V V = -
, ,
6 max
; g x P P = -
, , , ,
max
1 max max
, ;
g
h x F F l i = -
, , , ,
min
2 min min
, ;
g
h x F F l i = -
4. RESULTS AND DISCUSSIONS
The non-dominated sorting genetic algorithm-II (NSGA-II) [8]
is implemented to the present work. The algorithm involves the
single point crossover and the bitwise mutation for binary variables,
the simulated binary crossover (SBX) and the polynomial
mutation for real coded variables, and the parameter-less
approach is used to handle constraints.
For the optimization, SBX parameters namely the probability of
the crossover, the probability of the mutation, the crossover
distribution index, the mutation distribution index and basic seed
are assumed to be 0.1, 0.01, 5, and 10, 0.1 respectively, where
the number of real variables are 4 for the present problem. To have
better understanding of the behavior of design variables with
respective to objective functions in the final optimized population, a
population of 200 is taken with 1000 generations. Objective
functions are optimized at the operating point of magnetic
bearings. The Pareto front [6] is the locus of different non-
dominated solutions in the multi-objective space. The non-
dominated sorted genetic algorithm gives a group of solutions,
which are ranked according to the degree to which constraints
are satisfied and the objectives are optimized. Final solutions
optimize objective functions to different degrees. For a weighted
sum approach of the single objective optimization, it gives a single
optimum value for each run that may not be Pareto optimal. But
in the case of multi-objective optimization, we get different
Pareto optimum designs in a single run. From this group of
solutions the designer chooses proper design according to the
application that is intended.
There are two types of constraint handling, i.e., the hard and the
soft. In the case of hard, the violation of the constraint is not
tolerated; in the case of soft constraint the violation of the
constraint is tolerated to a specified degree In the present case,
the equality constraint of load to be supported in Equation (12) is
satisfied by determining the required current densities and flux
densities from Equations(16), 18) . All the other constraints in
Equations (17), (19), (20), are greater than or equal to type, and
all the constraints are treated to be hard.
Table 4. Initial Bounds of the design variables
Initial bounds
Variables :
c
r (m)
p
r (m)
c
l (m)
p
n
Lower limit: 0.3 0.03 0.2 4
Upper limit: 0.8 0.08 0.8 12
Table 5. Final Bounds of the design variables
Final bounds
Variables :
c
r
p
r
c
l
p
n
Lower limit: 0.306 0.0305 0.2 7
Upper limit: 0.78 0.0416 0.208 10
Table 6. Optimum bearing variables after final population.
Variables :
c
r
p
r
c
l
p
n
Optimum: 0.7109m 0.0351m 0.204 8
Table 7. Optimum dimensions of RMB
2
0.00167 ,
c
A m =
,
2
0.003398 ,
g
A m =
3
0.000833
s
V m =
3
0.000334 ,
c
V m =
6.474 ,
s
W Kg =
2.9615
c
W Kg =
The minimum and maximum values of the theoretical feasible
bounds on the design vector and final optimized solution are
given in Table 4 to 7 respectively. It could be noted that the
bounds obtained by GA are within the range of theoretical
bounds. From Table 5,the optimum ranges (in the final
population after 1000 generations) on the coil space radius, pole
tip radius, coil length; number of poles are 94.8%, 22.2%, 13.3%
and 12.5% of feasible ranges, respectively. The optimum ranges of
the pole tip radius, coil length are near the lower limit of the
feasible range and except the coil space radius.
International Journal of Computer Applications (0975 8887)
Volume 27 No.6, August 2011
24
0
2
4
6
8
10
12
14
1 124 247 370 493 616 739 862 985
Generat ions
W
e
i
g
h
t

o
f

R
M
B

(
K
g
)
Fig 3: Convergence graph for minimumweight of RMB
Convergence of the minimum weight with the generation is
plotted in Figure 3.
0
10
20
30
40
50
60
70
80
1 9 17 25 33 41 49 57 65 73 81 89 97
Iterations
A
v
e
r
a
g
e

C
u

l
o
s
s
(
W
)
Fig 4: Average power loss for 100 runs
8
8.5
9
9.5
10
10.5
1 10 19 28 37 46 55 64 73 82 91 100
Iterations
A
v
e
r
a
g
e

b
e
a
r
i
n
g

w
e
i
g
h
t
(
K
g
)
Fig 5: Average bearing weight for 100 runs
Results have been obtained with population size of 200 and 100
runs for 1000 generations. From figure 6 it is observed that the
decrease in the power-loss causes an equivalent increase in
the weight and visa versa.
Figure 6. Shows that optimum weight of bearing is 9.435 Kg
where as optimum copper loss is 54.635W.
9.4
9.45
9.5
9.55
9.6
9.65
9.7
9.75
9.8
40 50 60 70 80
Co p p er lo ss( W)
T
o
t
a
l

w
e
i
g
h
t

f
o
r

v
a
r
y
i
n
g

p
o
l
e
s

(
K
g
)
Wt =9 . 4 3 5 Kg( 9 2 . 5 N
Cu lo ss=5 4 . 6 3 5 W
Fig 6: copper loss versus bearing weight
0
0.0005
0.001
0.0015
0.002
0.0025
40 50 60 70 80
Copper loss(W)
C
o
i
l

a
r
e
a
(
m
2
)
A
B
C
Fig 7: Effect of coil area on copper loss
As shown in Figures 7, the cross sectional area of the coil
decrease with the power-loss to a minimumlimit, until the saturation
limit of the current density is reached. Between points B and C of
curve, solution points are scattered than between points A and B.
0
0.5
1
1.5
2
2.5
3
3.5
4
0 20 40 60 80
Copper loss(W)
W
e
i
g
h
t

o
f

c
o
i
l
(
K
g
)
A
B
C
Utopia pt.
Fig 8: (a) Effect of scattered region on bearing geometry.
Figure 8(a) shows procedure to select optimized bearing
geometry from the paretofront.Here we obtained copper loss
55W and weight of coil is 2.6 Kg.where as optimum values
obtained at run 93 for copper loss 54.63W, weight of coil 2.96Kg,
weight of stator 6.47Kg. Hence neglecting scattered region B-C
and applying pareto-front for region A-B, we get accuracy for
optimum values of bearing geometry as shown in figure 8(b).
When observed A-C, optimum design of the minimum
normalized distant member, an increase of 23 % in the power-
International Journal of Computer Applications (0975 8887)
Volume 27 No.6, August 2011
25
loss generated a decrease of 72. 2 % in the weight, where the
percentage of change is determined according to the formula for
Percent of change of quantity=
min
max min
100
X X
X X
| | -
=
]
-
\
Where,
X represents the quantity (i.e., the copper loss or weight)
considered.
X
max
and X
min
are the maximum and minimum values of the
corresponding quantities in the cases of minimum weight and
minimumpower-loss.
The weight of the coil versus copper loss shows similar trends as
the area of cross section before point B, but after point B the
decrease in weight with the Equation (7).
0
0.5
1
1.5
2
2.5
3
3.5
4
0 20 40 60 80
Copper loss(W)
W
e
i
g
h
t

o
f

c
o
i
l
(
K
g
)
A
B
C
Utopia pt.
(54.6W, 2.9Kg)
Fig 8 (b): Optimum bearing geometry using pareto-front.
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0 20 40 60 80
Copper loss(W)
A
r
e
a

o
f

g
a
p
(
m
2
)
A
B
C
Fig 9: Effect of area of gap on Copper loss
As shown in Figures 9, the copper loss increases with increase in
area of gap, until the saturation limit at point C of the current
density is reached. Between points B and C of curve, solution points
are scattered than between points A and B. Figure 10 shows
increase in coil area volume of coil increases which is in directly
proportional to weight of coil.
0
0.5
1
1.5
2
2.5
3
3.5
4
0 0.0005 0.001 0.0015 0.002 0.0025
Coil area(m
2
)
W
e
i
g
h
t

o
f

c
o
i
l
(
K
g
)
A
B
C
Fig 10: Effect of weight of coil on copper loss
200000
250000
300000
350000
400000
450000
500000
10 30 50 70
Copper loss(W)
C
u
r
r
e
n
t

d
e
n
s
i
t
y
,
(
A
/
m
2
)
A
B
Fig 11: Effect of current density on copper loss
Figure 11 shows that copper loss is accompanied by the increase
in the current density, which linearly increases up to its saturation
limit (i.e., point B) power-loss.
0
0.0002
0.0004
0.0006
0.0008
0.001
0.0012
0 0.001 0.002 0.003 0.004 0.005
Gap area(m
2
)
V
o
l
.
o
f

s
t
a
t
o
r
(
m 3
)
Fig 12: Effect of gap area on volume of stator
Figure 12 shows change in dimensions of pole affect pole face
area i.e. gap area. This means increase in pole size along axis of
rotor , increases pole face area or gap area due to which volume of
stator increases proportionally.
International Journal of Computer Applications (0975 8887)
Volume 27 No.6, August 2011
26
0
0 . 0 0 0 5
0 . 0 0 1
0 . 0 0 1 5
0 . 0 0 2
0 . 0 0 2 5
0 2 0 0 0 0 0 4 0 0 0 0 0 6 0 0 0 0 0
C u r r en t d en s it y ( A /m
2
)
C
o
i
l

A
r
e
a
(
m
2
)
A
B
C
Fig 13: Effect of coil area on current density
0
0.001
0.002
0.003
0.004
0.005
0 200000 400000 600000
Current densit y(A/m
2
)
G
a
p

a
r
e
a
(
m
2
)
B
C
A
Fig 14: Effect of gap area on current density
The constraint violation plots in the final population (i.e., after
1000 generations) are shown in Figure 13and 14.Constraints used
are greater than or equal to type, and as all the values are above
zero for all the solutions are feasible solutions. Curve A to B is for
unscattered solution and limiting values located at A and B are
feasible for finding optimum values of gap area and coil area using
pareto-front method.
0
0.2
0.4
0.6
0.8
1
0 0.05 0.1 0.15 0.2 0.25
Coil lengt h(m)
C
o
i
l

s
p
a
c
e

r
a
d
i
u
s
(
m
)
Fig 15: Coil length versus coil space radius
Figure 15 shows for final population after 100 runs, optimum
bounds for coil length and coil space radius are 0.2-0.21m and 0.3-
0.78 m.
Figure 16 shows all solution points are within pole limit 7 to 10.
Some scattered solution points in upper region but lower region is
dense with convex curve shape Fromfinal bound of poles 8 number
of poles are suitable for avoiding cross coupling effect in radial
magnetic bearing. With this curve we get optimum bearing weight
9.43 Kg for optimum number of 8 poles
9.4
9.45
9.5
9.55
9.6
9.65
9.7
9.75
9.8
0 2 4 6 8 10 12
Number of poles
W
e
i
g
h
t

o
f

A
M
B
(
K
g
)
Fig 16: Effect of varying pole on weight of bearing
0.3
0.31
0.32
0.33
0.34
0.35
0.36
0.37
0.38
0 0.0005 0.001 0.0015 0.002 0.0025
Coil area(m
2
)
M
a
g
n
e
t
i
c

f
l
u
x

d
e
n
s
i
t
y
(
T
)
Fig 17: Effect of Magnetic flux density on coil area
Figure 17 shows magnetic flux density increases with increase
in coil area.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
40 50 60 70 80
Copper loss(W)
M
a
g
n
e
t
i
c

f
l
u
x

d
e
n
s
i
t
y
(
T
)
Fig 18: Copper loss versus Magnetic flux density
Once the magnetic flux density is saturated as shown in Figure
18, the magnetic flux density is slightly decreases with increase in
the power-loss.
5. CONCLUSION
In this present work, the optimal design of radial active magnetic
bearings using MOGAs has been carried out. Objective functions
and associated constraints have been presented and discussed.
Two objectives have been considered, namely the minimizing
the copper loss and weight of bearing. Coil dimensions and
International Journal of Computer Applications (0975 8887)
Volume 27 No.6, August 2011
27
number of poles have been proposed as design variables based on
their dependency on objective functions. The geometries of the
designs with optimal individual objectives and a chosen design
from the optimal Pareto front based on the minimum normalized
distance criterion are compared. Performance analysis is carried
out graphically for bearing geometry and variables. The
behavior of final optimized population designs has also been
studied. Copper loss, weight of bearing, magnetic flux density and
coil current density are observed for different design parameters.
Moreover, in the final population after 1000 generations, the
optimum range of variables on the coil space radius, pole tip
radius, coil length, number of poles are 94.8%, 22.2%, 13.3% and
12.5% of feasible ranges, respectively. The optimum ranges of the
pole tip radius, coil length are near the lower limit of the feasible
range and that of the coil space radius is near the upper limit. In
optimum design of the minimum normalized distant member, an
increase of 23 % in the power-loss generated a decrease of 72.
2 % in the coil weight.
From graphs it is observed that objective functions are mutually
conflictive i.e. with the decrease in the power-loss causes an
equivalent increase in the weight and visa versa.Otimum value of
objective functions are 54.63 W and 9.43Kg.
6. ACKNOWLEDGEMENT
Corresponding author would like to thank Prof.R.V.chalam for
his valuable guidance for this research in organization National
Institute of technology, Warangal, India.
7. REFERENCES
[1] Bakay L., and Dubois M., Losses in an optimized 8-pole
radial magnetic bearing for long term flywheel energy
storage, IEEE transaction on magnetic, Canada, 2007,3-5.
[2] Bakay L., Dubois M., and Ruel J, Mass-Loss relationship
optimized 8-pole AMB for long term flywheel energy
storage, IEEE AFRICON, Kenya, 2009,.2-3.
[3] V.D. Bloodgood, Jr., N.J. Groom, C.P. Britcher, Further
Development of an Optimal Design Approach Applied to
Axial Magnetic Bearings, NASA-2000-7ismb-vdb.
[4] H. Chang, S.C. Chung, Integrated Design of Radial Active
Magnetic Bearing Systems using Genetic Algorithms,
Mechatronics, vol. 12(1), 2002, 19-36.
[5] S. Ghanmi, M. Guedri, M.-L. Bouazizi, N. Bouhaddi, Use
of Metamodels in the Multi-Objective Optimization of
Mechanical Structures with Uncertainties, International
Journal for Computational Methods in Engineering Science
and Mechanics, vol. 8(5), 2007, 283-302.
[6] Jagu S.Rao, Rajiv Tivari, Optimum design and analysis of
thrust magnetic beating using genetic Algorithm,
International Journal for Computational Methods in
Engineering Science and Mechanics, vol. 9, 2008, 223-245.
[7] C.A.C. Coello, An Updated Survey of GA-Based
Multiobjective Optimization Techniques, ACM
Computing Surveys, vol.32 (2), pp.109-143, 2000.
[8] K. Deb, A. Pratap, S. Agarwal, T. Meyarivan, A Fast and
Elitist Multiobjective Genetic Algorithm, NSGA-II, IEEE
Trans. Evol. Comput, vol. 6, 2002, 182-197.
[9] C.M. Fonseca, P.J. Fleming, An Overview of Evolutionary
Algorithms in Multi-objective Optimization, Evolutionary
Computation, Springer, vol. 3(1), 1995, 1-16.
[10] L.T. Bui, J. Branke, H.A. Abbas, Multiobjective
Optimization for Dynamic Environments, IEEE Paper
No. 0-7803-9363-5/05, 2005, 2349-2356.
[11] Y. GAO, L. Shi, P. Yao, Study on Multi-Objective
Genetic Algorithm, Proc.3rd World Cong. On intelligent
Control and Automation, 2000, 646-650.
[12] X. Hu, Z. Huang, Z. Wang, Hybridization of the Multi-
Objective Evolutionary Algorithms and the Gradient-based
Algorithms, IEEE Paper No.0-7803-7804-0 /03, 2003,
870-877.
[13] B.R.Rao, R.Tiwari, Optimum design of rolling element
bearing using genetic algorithm, Mechanism and machine
theory, 2007, 233-250.
[14] C.L. Malone, Power to Weight Optimization for
Magnetic Bearings, Magnetic Bearings, Magnetic Drives
and Dry Gas Seals Conference & Exhibition, Alexandria,
Virginia, July, 1993.
[15] Ying Gao, Lei Shi, Pinging Yao, Study on Multiobjective
Genetic Algorithm, IEEE transaction on Intelligent
control and Automation, China, 2000, 646-650.
[16] R.T.marler, J.S.Arora, Survey of multiobjective
optimization methods for engineering, Springer,Vol
26,2004,369-395.
[17] Tomoharu Nakashima, Hisao Ishibuchi, Genetic
algorithm based approaches for finding the minimum
reference set for nearest neighbor
classification,IEEE,1998,709-714.
[18] Deepti Chafekar, Liang shi, Jiang Xuan, multiobjective
genetic algorithm optimization using reduced models,
IEEE transaction on system, Vol 35(2), 2005, 261-265.
[19] Santosh shelke, R.V.Chalam, Optimum energy loss in
electromagnetic bearing, IEEE, vol.3, 2011,.374-378.
[20] E.H.Maslen, Text book on Magnetic Bearings,
University of Virginia, 36(4), 2000, 1009-1013.
[21] Santosh shelke, R.V.Chalam, Optimum copper loss
analysis of eight pole radial magnetic bearing: Multi
objective genetic algorithm, accepted for Springer,
Proceeding of International conf. ccpe 2011.